We have found that mtDNA depletion of tissues and cultured fibroblasts from patients with POLG1 mutations is associated particularly with mutations in catalytic domains. We have identified mosaic mtDNA depletion of mtDNA in primary fibroblast cultures from some patients with Alpers disease. In patients A–C, mtDNA was depleted in cells with reduced COX activity. These cells also had reduced mitochondrial membrane potential, suggesting that the depletion and not the point mutations underlie the major part of the respiratory chain defect in fibroblasts from these POLG1 patients (Supplementary Material, Fig. S4
). In all cases, cellular mosaic depletion was associated with two mutations in catalytic regions of POLG1, either in the polymerase or exonuclease domain or in one of each. The mtDNA depletion manifested primarily as a reduced number of nucleoids in the depleted cells and reflected the severity of the clinical and tissue phenotypes. The cellular mtDNA content may be an indicator of the underlying molecular mechanism that links genotype to phenotype. These data from a human disease suggest that eukaryotic DNA polymerases require both polymerase and 3′–5′exonuclease activities within the same molecule.
Previous studies of human POLG1 mutations have found little in the way of clear genotype–phenotype correlations, and these may have been more apparent in the present study as it has been possible to link a range of clinical phenotypes with laboratory data that included both tissue mtDNA content and cellular phenotypes. Our clinical findings were similar to previous studies, confirming that epilepsy and movement disorders (27
) are major features of POLG1 syndromes. Relatively complete information was available for the four patients who manifested mosaic mtDNA depletion, whereas the clinical data available on the other patients was largely retrospective, hence dependent on the local neurologist. We found that the cellular phenotype correlated better with the location of the mutation and the severity of the tissue depletion than with specific clinical features. Nevertheless, mosaic mtDNA depletion was evident in fibroblast cultures from only the most severely affected children, all of whom had died before the age of 16 months. Patient D had two mutations, R232G and T251I, in the exonuclease domain, the latter being in cis
with the P587L linker mutation. The P587L mutation usually co-segregates with T251I, but has occurred with 251T in a patient with a relatively mild phenotype (28
). While this suggests that P587L is mildly pathogenic, T251I may exacerbate its effect. Patient M was homozygous for the L304R exonuclease mutation and had a relatively mild phenotype, but neither liver nor fibroblasts were available for study. In contrast to the patients with two catalytic mutations, no patient who only had missense linker mutations was dead by 16 months or had mosaic cellular or tissue mtDNA depletion.
Depletion of mtDNA was present in liver in all cases where it was available for measurement, but was much less consistent than in muscle. All patients with hepatic mtDNA depletion had at least one catalytic mutation, whereas patients with purely linker mutations had little in the way of liver features. These were usually missense mutations, but one patient had a nonsense mutation in the polymerase domain resulting in mono-allelic expression (21
). We found cellular mtDNA depletion only in individuals with clear tissue depletion (Table ). The severity of the depletion was commensurate with the severity of the defect in replication that we have seen, and this is supported by the published literature (29
). These genotype–phenotype correlations may not have been seen in some previous studies because the diagnosis of mtDNA depletion syndrome is frequently imprecise. This is for both technical reasons (25
) and because of the lack of established age-adjusted normal ranges, and may explain some of the inconsistencies in the published literature. However, age-related changes in mtDNA content are undoubtedly also a feature of mtDNA depletion (18
). Furthermore, it is clear that the effect of specific mutations may be modulated by other missense mutations in cis
Summary of mutation by location and phenotypic severity: Colour represents severity running from dark orange (most severe) to white (least severe)
An important aspect of this study is the use of PicoGreen to visualize mtDNA packaged into nucleoids. While PicoGreen is less sensitive that antiDNA antibodies and does not detect nucleoids containing the lowest levels of mtDNA, it can be used more quantitatively than antiDNA antibodies, being sensitive to the nucleoid DNA content (23
). With this method, it could be shown that both the number of nucleoids and the average mtDNA content in individual cells declined with time in fibroblast cultures from the patients with the mosaic cellular depletion phenotype (Fig. ). During this decline, the distribution of nucleoid intensities remained similar within the range where the PicoGreen signal is quantitative (23
). This might indicate that the balance between nucleoid fusion and division depends more on the mtDNA content of the nucleoid than the number of nucleoids in the cell. It may also explain our observation that occasional mitochondria within TFAM depleted cells retained Mitotracker labelling (Supplementary Material, Fig. S2
). Occasional nucleoids of relatively normal mtDNA content in cells that are profoundly depleted of mtDNA may underlie the co-localization of Mitotracker and residual TFAM signal.
Like many replicases, POLG1 has a 3′ to 5′ nuclease activity that is used to excise incorrectly paired bases. Mutations in this domain have been seen in both childhood and adult onset POLG1 disease. In the mutator mouse model, POLG1 with deficient exonuclease activity impairs both polymerase fidelity and shortens lifespan (34
). By analogy with other DNA polymerases, the enzyme probably alternates between polymerizing and editing modes, as determined by competition between the two active sites for the 3′ primer end of the DNA (37
). Once the correct base has been incorporated, the enzyme moves forward 1 bp further, ready for the next precursor nucleotide to enter. Patient B, who is a compound heterozygote with one mutation in the polymerase and one in the exonuclease domain, has a severe cellular and clinical phenotype, comparable with patients A, C and D, who either have two polymerase (patients A and C) or two exonuclease mutations (patient D). The apparent lack of complementation between the POLG1 mutations in patient B supports the molecular model, by suggesting that normal mtDNA replication requires both polymerase and exonuclease activity within the same alpha subunit.
Previous studies of patients with POLG1 mutations have found limited correlation between phenotype and genotype (38
). Horvath et al
) reported that patients with Alpers disease generally had at least one catalytic mutation, most commonly in the polymerase domain, and that patients whose mutations both lay in the linker domain presented later. Bindoff and coworkers (27
) studied 26 individuals with two common linker mutations, A467T and W748S. The A467T mutation influences the interaction of POLG1 with POLG2 and the W748S mutation reduces both processivity and DNA binding. These authors found major differences in survival depending on genotype, with compound heterozygotes having a significantly shorter survival time than patients homozygous either for the A467T or W748S mutations. As the active form of POLG is known to be a compound heterotrimer containing one alpha and two beta subunits, this suggests a quaternary interaction between catalytic subunits in different heterotrimers. While our data set is much too small to draw parallel inferences, we saw nothing to suggest quaternary interactions between polymerase domains. Patient C, who was homozygous for the R1096C mutation, had a comparable cellular and clinical phenotype to patient A, who was a compound heterozygote for this mutation and T914P.
Six of the other POLG1 mutations we identified have not been reported previously. These comprise one nonsense mutation (R374X), one 10 amino acid deletion (W347_L356del), three simple missense mutations (H277L, R417T, H569Q) and one allele with two missense mutations (P587L;P589L). While P589L is novel, it is in cis with P587L which has been reported as a recessive mutation. H277L, R417T and H569Q are all within nine amino acids of previously reported recessive mutations and alter residues that are highly conserved across mammalian species. In addition, R417T may lead to aberrant splicing as the last nucleotide of exon 6 is mutated (1250G>C).
Wanrooij et al
) used two dimensional DNA analysis to show that cells, over-expressing POLG1 or TWINKLE containing catalytic defects, have both cellular depletion of mtDNA and replisome stalling. In some of their cell lines, the mtDNA depletion was apparent as reduced numbers of nucleoids. We have previously shown that intra-mitochondrial nucleotide imbalance occurs in cells from patients with mutations in TK2, dGK, TWINKLE and, potentially, TP (24
). Nucleotide imbalance is another potential cause of replisome pausing and depletion, and cells that are deficient in dGK may also have mosaic mtDNA depletion (25
). Replisome pausing may thus underlie mtDNA depletion due to either POLG1 mutations (41
) or to nucleotide imbalances (43
). The mosaic cellular depletion that we have demonstrated may thus be a manifestation of severe replication stalling. As in the model proposed by Wanrooij et al
., we found that each of the fibroblast cultures from patients A–D (that had the most severe cellular depletion) had at least two POLG1 mutations lying in catalytic domains. The depletion in the patient cells was less pronounced than in the in vitro
study, however, the progressive depletion observed with successive passages of patient cells may partly recapitulate the depletion observed on induction of the mutant POLG1 in the experimental study.
The clinical findings in this group of patients are similar to those found in previous studies of POLG1 mutations in Alpers disease. The prominence of epilepsy has been highlighted before, as well as the variability in pointers to mitochondrial disease. Movement disorders appear to be very much more common than in maternally inherited mtDNA disease, especially if this feature is broadened to include epilepsia partialis continua. Thus, epilepsia partialis continua and movement disorders may provide an important pointer to POLG1 mutations. In one case, fatal deterioration followed valproate therapy and others have reported a clear improvement on stopping therapy with this anticonvulsant. However, several other patients, with a generally milder phenotype, have been treated with this drug for short periods without apparent problem. The relationship between the hepatic encephalopathy associated with valproate therapy and pre-existing POLG mutations (carrier frequency <1 in 150) is worthy of further study.
Among a group of patients with POLG1 mutations, we have identified and characterized mosaic cellular depletion of mitochondrial DNA in four patients, each with two mutations affecting catalytic domains. These mutations appear to have severe consequences, both for the clinical phenotype and the mtDNA content of affected tissues. Patients with demonstrable tissue mtDNA depletion always had at least one catalytic domain or nonsense mutation. Our findings add general support to the current model of the role of POLG1 in the replisome.